rel-ratebeer RateBeer beer reviews database
Database Description: Founded in the early 2000s, RateBeer was one of the most popular beer review platforms, playing a central role in the craft beer community by enabling users to rate and review beers, discover new breweries, and share recommendations. The dataset spans more than two decades of user preferences with beer-related entities. It is well-structured with separate tables for beers, places, users, and brewers, all linked through well-defined foreign keys. Many entity tables are attribute-rich, and interaction tables capture multi-aspect feedback, such as sub-scores for flavor, aroma, and ambiance, along with detailed textual reviews.
Database Statistics:
| Num of Tables | 13 |
| Num of Rows | 13,787,005 |
| Num of Columns | 221 |
| Starting Time | 2000-04-02 |
| Validation timestamp | 2018-09-01 |
| Testing timestamp | 2020-01-01 |
| Time window | 3 months |
Database schema:
To load this relational database in RelBench, do:
from relbench.datasets import get_dataset
dataset = get_dataset("rel-ratebeer")
References:
Dataset License: CC BY-NC-SA 4.0.
Acknowledgements: We thank Joseph Tucker from RateBeer for assisting in facilitating access to this dataset.
Entity Classification Tasks
beer-churn
Task Description: For each beer, predict if it will receive zero ratings in the next 90 days.
Evaluation metric: AUROC
user-churn
Task Description: For each active user, predict if they will rate zero beers in the next 90 days.
Evaluation metric: AUROC
brewer-dormant
Task Description: For each brewer, predict if it will release zero beers in the next year (risk of going dormant).
Evaluation metric: AUROC
Entity Regression Tasks
user-count
Task Description: Predict the number of ratings a user will give in the next 90 days.
Evaluation metric: MAE
beer_ratings-total_score
Task Description: For each user, given a beer, predict the total score rating the user will give to the beer.
Evaluation metric: MAE
Link Prediction Tasks
user-beer-favorite
Task Description: For each user, predict the top 10 beers they will next add to their Favorites list.
Evaluation metric: MAP
user-beer-liked
Task Description: For each user, predict the top 10 beers they will rate at least 4.0 / 5.0.
Evaluation metric: MAP
user-place-liked
Task Description: For each user, predict the top 10 places they will rate at least 80 / 100.
Evaluation metric: MAP